Mobile Phone Data for Society and Privacy for the Individual: From the Conflict to a Synergy in Transport Flows Analysis

Smart city demands sound knowledge on citizen presence and mobility. Mobile phone data is the source of such knowledge. Based on the cellular data, urban population and its mobility can be mapped to spatial resolution of tens of meters and temporal resolution of minutes, and the precision will soon rise to meters and seconds. However, increasing quality and precision of location data comes at the expense of an unprecedented drop in personal privacy. Cellular data suppliers clearly understand this danger and are stalled in a controversial tradeoff. As a result, practical implementation of a vast spectrum of smart city ideas is delayed again and again.

Transportation and urban science define demands and impose essential limitations concerning the level of aggregation of cellular data. For example, to establish a new bus stop or add a dedicated bus lane to the existing road, the locations and routes of urban travelers should be known at high spatial and temporal resolutions. The maps of transportation flows, by modes (private cars, public transport, bikers, pedestrians), translated into spatially extended graphs, are the major information component for the modern methods of transportation planning and management. Systematic reassessment of urban traffic flows is cornerstone of the future smart city that will evolve with respect to evolving citizens’ demands.

Privacy vulnerabilities are inherent for the location-based data and hiding personal identifiers (e.g., by replacing them with pseudonyms) is insufficient to guarantee anonymity since the location could still lead to the identification of the individual. Relevant security and decision-theoretic methods are thus needed for the anonymization of individual location-based queries in order to preserve privacy when extracting value for the society from locational data.

Our research aims at establishing the necessary and sufficient space-time resolution and level of aggregation of mobility data that are required for the smart city transportation planning and management and, at the same time, clearly understand the potential harms to privacy, avoid disclosure of individual information and establish the forms of mobility data supply and procedures of data management and analysis in order to guarantee individual privacy. The outcome of this interaction should be clear privacy-preserving rules of mobility data aggregation for transportation planning and management.

To achieve this goal, we will:

• Develop privacy-preserving mathematical models for the dynamics of co-trajectories using clustering methods and continuous-time Markov chains;
• Train the model with data from millions of transformed individual trajectories over several months using latest distributed fault-tolerant big-data algorithms;
• Quantify the effect of privacy, that is specified through the spatial resolution of individual trajectories, upon the statistical risk of the model’s parameter estimates with respect to the requirements of transportation planning and management.

The empirical research will be conducted in the selected area of the Tel Aviv metropolitan region.